Long-term history of vehicle collisions on the endangered Nēnē (Branta sandvicensis)
Long-term history of vehicle collisions on the endangered N?en?e (Branta sandvicensis)
Christopher A. LepczykID 0 3
Jean E. Fantle-LepczykID 0 1 3
Kathleen Misajon 3
Darcy Hu 2 3
David C. Duffy 3
0 School of Forestry and Wildlife Sciences, Auburn University , Auburn, AL , United States of America
1 Department of Zoology, University of Hawai'i at Ma?noa, Honolulu, HI, United States of America, 3 U.S. National Park Service, Hawai'i Volcanoes National Park , Hawai'i National Park, HI , United States of America
2 U.S. National Park Service , Pacific West Regional Office, Hawai'i National Park , HI, United States of America, 5 Pacific Cooperative Studies Unit, Department of Botany, University of Hawai'i at Ma?noa , Honolulu, HI , United States of America
3 Editor: Bi-Song Yue, Sichuan University , CHINA
Funding: Funding for this work was provided by
the NPS to the University of Hawai?i through the
Hawai?i-Pacific Islands Cooperative Ecosystem
Studies Unit, Task Agreement No. P14AC01740
with CAL and DCD. Sponsors collected and
analyzed the nene data, assisted in design of
project and preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Millions of birds in the United States die annually due to vehicle collisions on roads.
Collisions may be of particular interest for species of conservation concern, such as the
endangered Hawaiian goose (Ne? n e?), which is endemic to Hawai?i. Using a nearly 40-year dataset
of Ne? n e? road mortality in and around Hawai?i Volcanoes National Park, we sought to answer
the following research questions: 1) has Ne? n e? mortality changed over time? 2) are there
times of the year in which mortality is greatest and does it relate to specific events in the
species? lifecycle? 3) does age at mortality differ over time, space, or sex? 4) given that existing
mortalities appear to occur only in certain locations, do the number of mortality events differ
across these locations; 5) does mortality rate show any density dependence? and, 6) are
mortality rates related to numbers of visitors or vehicles? Between 1977 and 2014, a total of
92 Ne? ne? died from vehicle collisions; while absolute mortality increased over this time, the
mortality rate remained the same. Similarly, average age of mortality increased over time,
but did not differ by location or sex. Between 1995 and 2014, Ne? n e? population size and
mortality rates were not correlated. Mortality was greatest in November and December
(breeding season) and lowest in June. Most of the mortality occurred along just three stretches of
road in and around the park, with the number of mortalities split about evenly inside and
outside of the park. Furthermore, Ne? n e? mortality was unrelated to the number of visitors or
traffic volume in the park. These findings suggest vehicle collisions are a growing concern for
Ne? n e?, but that management actions to reduce mortality can be targeted at specific road
segments and times of the year.
Anthropogenic sources of mortality represent a significant loss to many bird species across
North America [
]. Direct sources of human-caused mortality range from window collisions
to feral cats [
]. Vehicle collisions are a growing concern among causes of direct
anthropogenic mortality [
]. Although bird-vehicle collisions resulting in injury or death of birds
occur on roads [
], not all species of birds suffer negative population repercussions from roads
Recent estimates indicate that between 89 and 340 million birds in the United States and
] and ~9 to ~19 million birds in Canada [
] die each year due to vehicle collisions on roads.
While vehicle-caused bird mortality differs by taxonomic group and season [
about mortality rates in terms of regional, seasonal, and taxonomic patterns is still somewhat
]. However, several attributes have been correlated to the probability of vehicle mortality
across species. For instance, bird species with a large brain relative to their body size tended to
avoid vehicles more often than species that had a small brain [
]. Likewise, birds that forage,
nest, or roost near roads have a greater likelihood of collision [
For species of conservation concern, vehicle collisions may be one more factor hindering
recovery efforts. In fact, three endangered bird species in the US are considered to have
population level impacts due to vehicle collisions: The Florida Scrub-Jay (Aphelocoma coerulescens),
Audubon?s Crested Caracara (Polyborus plancus audubonii), and the Hawaiian Goose or Ne?ne?
(Branta sandvicensis; [
]). Amongst these three species, one of particular concern is the Ne?ne?,
which is endemic to the Hawaiian Islands and faces a host of other threats (e.g., introduced
mammalian predators, habitat loss; [
]). Originally occurring across a number of the islands in
the archipelago, the population was estimated at over 25,000 in the late 1700s but was reduced
to less than 50 individuals by the 1940s [
]. During the 1950s, captive breeding was initiated to
begin recovery of the goose. Currently the statewide population is estimated at over 3,000
individuals , but the species still faces many challenges to recovery. A survey of 300
opportunistically collected Ne?ne? carcasses in Hawai?i found that vehicle collisions accounted for 5.7% of all
deaths where the cause of mortality could be identified . Notably, these mortality figures are
unlikely to be representative of overall mortality rates because they were based on opportunistic
sampling, and carcasses from vehicle mortalities are rarely sent in for necropsy as the source of
mortality is known. In a recent assessment of resighting and carcass collection data at Hawai?i
Volcanoes National Park (HAVO), where approximately 90% of the park?s population is
banded and resighting effort is high, vehicle collisions accounted for 14% of mortality from
2009?2016 (K. Misajon, unpublished data). Likewise, a previous assessment of mortality factors
indicated that vehicle collisions were the leading cause of death [
]. Ne?ne? are amongst the
most endangered waterfowl in the world [
] and are a long-lived species with relatively low
recruitment rates at HAVO [
]. Thus, increased adult mortality may have serious implications
for the population. Evaluating and understanding causes of mortality are critical for developing
and improving management methods that can contribute to species recovery .
To aid our understanding of Ne?ne? mortality and help reduce vehicle strikes, we examined
the trends and circumstances of vehicle collisions. Specifically, we sought to answer six basic
questions. First, has Ne?ne? mortality due to vehicles changed over time and space over the past
40 years? Second, are there times of the year in which vehicle-caused mortality is greatest and
does it relate to specific events in the species? lifecycle? Third, does age at mortality differ over
time, space, or sex? Fourth, given that existing mortalities appear to occur only in certain
locations, do the number of mortality events differ across these locations? Fifth, does the
vehiclecaused mortality rate show any density dependence? Sixth, are mortality rates related to
numbers of visitors or vehicles?
2 / 11
The US National Park Service has kept records of Ne?ne? road mortality events from 1977 to
2014 in and adjacent to HAVO (Fig 1) on the island of Hawai?i. Road mortalities were reported
to HAVO both by the general public and park staff year round, as the species is non-migratory.
These mortalities were then relayed to a biologist, who followed up to verify and document the
event. While records of vehicle-related mortality likely are incomplete (e.g., they could miss
birds that were hit and unreported), they likely represent most of the vehicle strike deaths
within the study area for several reasons. Specifically, Ne?ne? are well-known in Hawai?i (e.g.,
they are the State Bird), the park?s Ne?ne? restoration work is familiar to the community by
virtue of periodic articles in the local media, the national park is a visible and well-known
neighbor and easy point-of-contact for residents and staff commuting through the study area, park
staff receive regular information about Ne?ne? management and are familiar with the park Ne?ne?
biologist, and usually there is someone available to respond quickly to reports of dead geese.
Thus, we believe the data are an unbiased sample of Ne?ne? killed by vehicles over the course of
the study. The park also tallies numbers of visitors and, since 1990, numbers of vehicles
entering through its two entry points.
From the long-term road mortality data in and around HAVO, we quantified the total
number of Ne?ne? killed by vehicles between 1977 and 2014 by year, age, sex, location, and
month. In the case of age, we classified goslings as 0.5 years old for analysis and used the
minimum age of individuals in cases where a bird could be older. Both mortality and age data were
Fig 1. Hawai?i Volcanoes National Park (HAVO) study region with major road segments.
3 / 11
log transformed to meet assumptions of normality for regression analysis. We evaluated if
mortality and age at mortality had changed over the nearly 40 years using simple linear
regression. Further, we evaluated if the number of individuals killed varied by age using simple linear
regression. We evaluated if age at mortality differed by location inside or outside of the park as
well as sex using a t-test. Similarly, we evaluated if age at mortality differed by specific road
segment using ANOVA. To determine if mortality differed by location, we compared mortalities
among the nine road segments, which ranged from 1.6 to 5.6 km in length, based on where the
mortalities were documented, and evaluated if total mortalities over time were similar by
location using a chi-square test. Likewise, we evaluated monthly totals to determine if road
mortality rates differed using a chi-square test. For years in which both Ne?ne? population estimates
and road mortality data were available (1995 through 2014) we calculated the percent of the
population killed by vehicles each year. We evaluated if mortality and population size were
related, as well as if the annual percent mortality changed over time, using simple linear
regression. Likewise, we evaluated if annual Ne?ne? mortality was related to the annual numbers of
vehicles or visitors using simple linear regression. We carried out all analyses in Systat 13, with
a P < 0.05 considered significant.
Between 1977 and 2014, 92 Ne?ne? were reported to have died from vehicle collisions in and
adjacent to HAVO, for an average of 2.42 ? 2.21 (SD) birds per year. Over this nearly 40-year
period, Ne?ne? vehicle collisions increased significantly (F1,28 = 6.96, p = 0.013, r2 = 0.20; Fig 2).
However, the average annual percent mortality rate remained unchanged over the years in
which population estimates were conducted (F1,16 = 0.08, p = 0.79). Mortality also varied
significantly by time of year (?2 = 40.26, df = 11, p < 0.001), with December and November being
the months of greatest mortality and June being the lowest (Fig 3). The average age at mortality
for Ne?ne? was 4.47 (range 0.5?20, n = 83) years old, with the number of mortalities decreasing
significantly by age (F1,15 = 29.23, p < 0.001, r2 = 0.66; Fig 4). Males (n = 33) and females
(n = 32) were killed in nearly equal numbers. The average age of Ne?ne? at the time of collision
Fig 2. Annual Ne?ne? mortality due to vehicles from 1974?2014.
4 / 11
Fig 3. Monthly Ne?ne? mortality summed over the 40-year period.
increased over time (F1,81 = 98.14, p = 0.025, r2 = 0.06; Fig 5). However, age at mortality was
unrelated to whether the collision occurred inside or outside of the park (t = -0.16, df = 81,
p = 0.88), the specific road segment (F8,74 = 12.09, p = 0.79), or the sex of the bird (t = 0.45,
df = 63, p = 0.66). The number of mortality events was split nearly evenly between roads inside
(n = 45) and outside (n = 47) the park. However, three of the nine road segments accounted
for 75% of the mortality events (Fig 6), resulting in significantly different numbers of
mortalities by road location (?2 = 84.67, df = 8, p < 0.001; Fig 7).
Fig 4. Age of Ne?ne? at time of mortality. Because not all goslings could be aged to exact month we considered them all 0.5
5 / 11
Fig 5. Age of Ne?ne? at time of mortality compared to year when mortality occurred.
Between 1995 and 2014, the Ne?ne? population increased significantly from approximately
150 individuals to 265 individuals (F1,16 = 17.17, p = 0.001). However, population size and
vehicle-related mortality rates were not correlated ((F1,17 = 0.68, p = 0.42; Fig 8). Although
visitor numbers slightly increased over time since the late 1970s, the increase was not significant
(F1,28 = 1.33, p = 0.26). Likewise, there was no relationship between the number of visitors in
the park and the number of mortality events in a given year (F1,28 = 0.69, p = 0.41). On the
other hand, since vehicle data collection began in 1990, the number of vehicles entering the
park at Gates 1 and 2 increased 50% (from 451,531 vehicles/yr to 679,055 vehicles/yr),
indicating a statistically significant rise in traffic (F1,20 = 14.15, p < 0.001). However, this rise in
vehicle volume was not correlated to the number of Ne?ne? mortalities (F1,20 = 1.39, p = 0.25).
Among our six research questions, we found support for three of them. Specifically, Ne?ne?
mortalities due to vehicle collisions increased over time. Second, Ne?ne? mortality varied within
years, with the greatest numbers occurring in December and November, corresponding to
their breeding season [
]. Third, Ne?ne? demonstrated spatial aggregation in mortality, with
the majority of collisions occurring on just three road segments. On the other hand, we found
no relationships between mortality and Ne?ne? population size, the number of park visitors, or
the number of vehicles entering the park. Likewise, while the average age at mortality rose over
6 / 11
Fig 6. Total Ne?ne? mortality by road segment within the park and location of the segment inside or outside of the park. Road segment
lengths (km) are denoted above each bar.
time, there were no relationships with age and specific road segments, location inside or
outside the park, or sex.
Though relatively few bird-vehicle collision studies have been conducted, our findings of
spatiotemporal hotspots in mortality are consistent with similar studies (e.g., [
Specifically, most of the mortalities are occurring in the months during breeding and fledging and
mainly along three road segments. While the number of Ne?ne?-vehicle mortalities increased
over the nearly 40 years investigated, during the last 20 years the annual rate of mortality did
not. That is, from 1995 onward, the percent of the population killed due to vehicles was
similar. The lack of correlation of mortality with bird population size, traffic volume, and number
of visitors suggests that there may not be a direct causal relationship with mortality.
Alternatively, the lack of correlation could simply be an artifact of a small sample size or random
chance events. Notably, while traffic volume has been related to mortality rates for some
animal species , for birds, no such correlations have been found in other studies [
Regardless, the rise in vehicle-caused deaths is troubling, as they could serve as an additive
source of mortality for a species already facing a host of other anthropogenically related
challenges to recovery .
Our findings have several important management implications for Ne?ne?. Management
actions can target the three road segments where most mortalities occurred (K?puka K?hali?i
to Muliwai a Pele, Ka?? Boundary, Namakanipaio to Peter Lee; (Fig 4), and also focus on the
months when the majority of collisions occurred (November to February, May). Lastly,
considering that about half of the deaths occurred outside HAVO?s management jurisdiction, on a
state highway, reducing collisions across the landscape requires state and federal agencies to
work together in developing and implementing conservation and recovery actions.
7 / 11
Fig 7. Locations of Ne?ne? vehicle strikes.
8 / 11
Based on investigations of Ne?ne? killed on roads and intensive observations of live birds
during the course of annual monitoring, researchers noted that geese mainly frequent roads either
to forage alongside and nearby, or to cross roads separating nesting sites from brooding
habitats, often with flightless goslings. In some locations, the trek between nest and brooding site
occurs once, shortly after hatching. However, in other locations, families may commute
regularly, or periodically, between daytime foraging areas on one side of a road and night roost
sites on the other. Strategies to reduce vehicle strikes may differ in these two circumstances,
but likely will include actions from each of two generalized approaches to reducing
wildlifevehicle collisions: those attempting to change animal behavior and those focusing on changing
driver behavior [
Signage is the most common way to change driver awareness and behavior. Though
relatively few studies have evaluated the effectiveness of warning signs [
], driver simulation
studies have demonstrated they raise driver awareness [
], and that when signs were erected,
they resulted in reduced collisions for deer in the year after installation [
]. Furthermore, the
use of temporary warning signs that had reflective flags and solar-powered flashing amber
lights resulted in a marked reduction in collisions and resulted in reduced traffic speeds [
Use of more dynamic imagery in warning signs also may improve driver awareness and
response times [
Additional approaches to changing driver behavior could be evaluated in key locations.
These include temporary digital signs that can display speed limit, verbal or iconographic
warnings and/or other information, and temporary digital speed feedback signs. Devices
such as rumble strips to slow traffic also could potentially alert Ne?ne? to oncoming vehicles.
However, reflectors designed to alert deer to vehicles appeared to have a diminishing effect
over time [
]. Because approaches to alert and/or slow drivers, including signage and
mechanical methods, have not been researched extensively regarding their effectiveness in
reducing wildlife mortality, such mitigation techniques should be carefully evaluated if
employed for Ne?ne?.
Among techniques focused on changing animal behavior, wildlife fencing is the single
most recommended approach to reduce vehicle deaths [
]. Another technique is
vegetation management, such as eliminating lush grass or other forage species on road
shoulders and edges, to reduce the attraction of herbivorous birds. Vegetation management
should occur concurrently with wildlife fencing to reduce bird activity within the road
corridor. This combined method is currently in use on Hawai?i Island along a one mile section
of the Saddle Road (State Route 200), and a subsequent reduction in vehicle collisions with
Ne?ne? has been observed (Joaquin Mello, pers. comm.). A third recommended approach is
the use of underpasses, in which Ne?ne? could traverse beneath the road bed [
fences and underpasses could be useful in areas where birds must cross on foot, such as
along segments separating nesting from brooding sites, this technique may not be feasible
in all areas. However, due to their widespread success and recommendation, small scale
testing within the park could be an initial step to inform further mitigation actions for the
species. In particular, before-after control impact (BACI) evaluation of the efficacy of any
mitigation actions, preferably over multiple years [
], can help determine if further
measures are needed.
As we have demonstrated, Ne?ne? mortality from vehicles has been increasing over time, but
has been occurring at times and locations that management can target in order to reduce
mortality. Studies such as ours may be of particular value in using patterns in vehicle mortality to
inform management actions in the context of the species? underlying biology and ecology.
Finally, our analysis of exceptionally long-term data on Ne?ne? road deaths helps address the
dearth of research on vehicle collisions with both non-game species and smaller animals.
9 / 11
S1 Table. Lepczyk et al. Nene vehicle collision data.
We wish to thank all of the individuals who recorded Ne?ne? mortality events, allowing our
analysis to be possible. We also thank Scott Kichman, NPS Inventory and Monitoring Program,
for his contributions to the maps, and Sarah Allen and two anonymous reviewers for
comments on the draft manuscript. Data were collected under Hawai?i Volcanoes National Park?s
Endangered Species Recovery Permit #TE-018078-20 (and previous versions).
Conceptualization: Christopher A. Lepczyk, Kathleen Misajon, Darcy Hu.
Data curation: Kathleen Misajon.
Formal analysis: Christopher A. Lepczyk.
Funding acquisition: Kathleen Misajon, Darcy Hu.
Investigation: Christopher A. Lepczyk, Kathleen Misajon.
Methodology: Christopher A. Lepczyk.
Project administration: Christopher A. Lepczyk, Kathleen Misajon, Darcy Hu, David C.
Writing ? original draft: Christopher A. Lepczyk, Jean E. Fantle-Lepczyk.
Writing ? review & editing: Christopher A. Lepczyk, Jean E. Fantle-Lepczyk, Kathleen
Misajon, Darcy Hu, David C. Duffy.
10 / 11
1. Loss SR , Will T , and Marra PP. Direct human-caused mortality of birds: improving quantification of magnitude and assessment of population impact . Frontiers in Ecology and the Environment . 2012 ; 10 : 357 - 364 .
2. Loss SR , Will T , and Marra PP. The impact of free-ranging domestic cats on wildlife of the United States . Nature Communications . 2013 ; 4 : 1396 . https://doi.org/10.1038/ncomms2380 PMID: 23360987
3. Erritzoe J , Mazgajski TD , and Rejt L. Bird casualties on European roads-a review . Acta Ornithologica . 2003 ; 38 : 77 - 93 .
4. Kociolek AV , Clevenger AP , St Clair CC , and Proppe DS . Effects of road networks on bird populations . Conservation Biology . 2011 ; 25 : 241 - 249 . https://doi.org/10.1111/j.1523- 1739 . 2010 . 01635 . x PMID : 21284729
5. van der Ree R , Jaeger JAG , van der Grift EA , and Clevenger AP . Effects of roads and traffic on wildlife populations and landscape function: road ecology is moving towards larger scales . Ecology and Society . 2011 ; 16 ( 1 ): 48 . [online] URL: http://www.ecologyandsociety.org/vol16/iss1/art48/
6. Fahrig L , and Rytwinski T. Effects of roads on animal abundance: an empirical review and synthesis . Ecology and Society . 2009 ; 14 ( 1 ): 21 . [online] URL: http://www.ecologyandsociety.org/vol14/iss1/ art21/
7. Loss SR , Will T , and Marra PP. Estimation of bird-vehicle collision mortality on U.S. roads . The Journal of Wildlife Management . 2014 ; 78 : 763 - 771 .
8. Calvert AM , Bishop CA , Elliot RD , Krebs EA , Kydd TM , Machtans CS , et al. A synthesis of humanrelated avian mortality in Canada . Avian Conservation and Ecology . 2013 ; 8 ( 2 ): 11 . http://dx.doi.org/10. 5751/ACE-00581-080211
9. Husby A , and Husby M. Interspecific analysis of vehicle avoidance behavior in birds . Behavioral Ecology . 2014 ; 25 : 504 - 508 .
10. Huijser MP , McGowen P , Fuller J , Hardy A , Kociolek A , Clevenger AP , et al. Wildlife-vehicle collision reduction study . Report to Congress. U.S. Department of Transportation, Federal Highway Administration, Washington, D.C. , USA. 2007 .
11. Banko PC , Black JM , and Banko WE . Hawaiian Goose (Nene) (Branta sandvicensis) . The Birds of North America. 1999; Number 434 .
12. Baldwin PH . The Hawaiian Goose, its distribution and reduction in numbers . Condor . 1945 ; 47 : 27 - 37 .
Nene Recovery Action Group. State-wide population estimate for nene . 2016 .
Work TM , Dagenais J , Rameyer R , and Breeden R. Mortality patterns in endangered Hawaiian geese (Nene; Branta sandvicensis) . Journal of Wildlife Diseases . 2015 ; 51 : 688 - 695 . https://doi.org/10.7589/ 2014-11-256 PMID: 26161721
15. Rave EH , Cooper A , Hu D , Swift R , and Misajon K. Population and reproductive trends of Nene (Branta sandvicensis ) in Hawaii Volcanoes National Park , 1989 - 1999 . Wildfowl. 2005 ; 55 : 7 - 16 .
16. Black JM . Threatened waterfowl: recovery priorities and reintroduction potential with special reference to the Hawaiian Goose . In: Marzluff JM , Sallabanks R , editors. Avian conservation: research and management . Washington, D.C: Island Press; 1998 . pp. 125 - 140 .
17. Husby M. Factors affecting road mortality in birds . Ornis Fennica . 2016 ; 93 : 212 - 224 .
18. Canal D , Camacho C , Mart??n B, de Lucas M , Ferrer M . Magnitude, composition and spatiotemporal patterns of vertebrate roadkill at regional scales: a study in southern Spain . Animal Biodiversity and Conservation . 2018 ; 41 .2: 281 - 300 .
19. Gunson KE , Mountrakis G , and Quackenbush LJ . Spatial wildlife-vehicle collision models: A review of current work and its application to transportation mitigation projects . Journal of Environmental Management . 2011 ; 92 : 1074 - 1082 . https://doi.org/10.1016/j.jenvman. 2010 . 11 .027 PMID: 21190788
20. Clevenger AP , Chruszcz B , and Gunson KE . Spatial patterns and factors influencing small vertebrate fauna road-kill aggregations . Biological Conservation . 2003 ; 109 : 15 - 26 .
21. Black JM . The Nene Branta sandvicensis Recovery Initiative: research against extinction . Ibis . 1995 ; 137 : S153 - Sl60 .
22. Found R , and Boyce MS. Warning signs mitigate deer-vehicle collisions in an urban area . Wildlife Society Bulletin . 2011 ; 35 ( S1 ): 291 - 295 .
23. Ja ?gerbranda AK, and Antonson H. Driving behaviour responses to a moose encounter, automatic speed camera, wildlife warning sign and radio message determined in a factorial simulator study . Accident Analysis & Prevention . 2016 ; 86 : 229 - 238 .
24. Sullivan TL , Williams AF , Messmer TA , Hellinga LA , and Kyrychenko SY . Effectiveness of temporary warning signs in reducing deer-vehicle collisions during mule deer migrations . Wildlife Society Bulletin . 2004 ; 32 : 907 - 915 .
25. Cian L , Krishna A , and Elder RS . A sign of things to come: behavioral change through dynamic iconography . Journal of Consumer Research . 2015 ; 41 : 1426 - 1446 .
26. Ujvari M , Baagoe HJ , and Madsen AB . Effectiveness of wildlife warning reflectors in reducing deer-vehicle collisions: A behavioral study . Journal of Wildlife Management . 1998 ; 62 : 1094 - 1099 .
27. Bennett AF . Roads, roadsides and wildlife conservation: a review . In: Saunders DA , Hobbs RJ , editors. Nature Conservation 2 : The role of corridors . Surrey Beatty and Sons Pty Ltd; 1991 . pp. 99 - 118 .
28. Rytwinski T , Soanes K , Jaeger JAG , Fahrig L , Findlay CS , Houlahan J , et al. How effective is road mitigation at reducing road-kill? A meta-analysis . PLoS ONE . 2016 ; 11 ( 11 ): e0166941. https://doi.org/10. 1371/journal.pone. 0166941 PMID: 27870889
29. Huijser MP , McGowen P , Clevenger AP , and Ament R. Wildlife-vehicle collision reduction study: best practices manual . Report to Congress. U.S. Department of Transportation, Federal Highway Administration, Washington, D.C. , USA. 2008 .